Recent years have seen the development of biometric authentication techniques for authenticating a user of an apparatus or system based on a biometric image representing the user's biometric information such as a hand or finger vein pattern, a fingerprint or a palm print. A biometric authentication device using such biometric authentication techniques acquires, as an input biometric image, a biometric image representing, for example, the biometric information of the user who is to use the biometric authentication device. Then, the biometric authentication device matches the user's input biometric information represented in the input biometric image with registered biometric information which is biometric information represented in the preregistered images of registered users. If it is determined as a result of the matching process that the input biometric information matches the registered biometric information, the biometric authentication device authenticates the user as being a legitimate registered user. Then, the biometric authentication device permits the authenticated user to use a device including the biometric authentication device or some other device connected to the biometric authentication device.
In such a biometric authentication device, higher authentication accuracy is preferred so as to not inconvenience a user. Thus, a technique using fingerprints of two or more fingers for biometric authentication using the fingerprints as biometric information in order to improve the authentication accuracy (for example, see Japanese Patent Laid-Open No. 2001-92961 and Published Japanese Translation of PCT International Publication for Patent Application No. 1999-514771) is suggested.
Biometric authentication techniques include a so-called 1:1 authentic mode and a 1:N authentic mode. In the 1:1 authentic mode, by inputting user's identification information, such as a user name or an identification number, together with user's input biometric information, registered biometric information corresponding to the identification information is identified. Then, a biometric authentication device matches the input biometric information only with the identified registered biometric information.
In contrast, in the 1:N authentic mode, user's identification information is not input into a biometric authentication device, so that registered biometric information corresponding to the user may not be identified. Therefore, the biometric authentication device matches input biometric information with all preregistered biometric information. Then, the biometric authentication device authenticates the user as a registered user corresponding to the registered biometric information that best matches the input biometric information. Accordingly, with increasing the number of registered users, that is, with increasing the amount of the registered biometric information, the number of times of execution of a matching process is also increased. Therefore, time necessary for a biometric authentication process is also inevitably prolonged. When a biometric authentication technique is utilized particularly in a large-scale system, the number of registered users may be very high. There has been a fear that, when a server is used in a large-scale client system, for executing a biometric authentication process, access concentrates on a server to increase a load on the server, so that latency time from inputting of biometric information by a user to obtaining of an authentication result is prolonged.
However, shorter time necessary for a biometric authentication process is preferred from the viewpoint of user's usability. Thus, a technique for limiting registered biometric information utilized in matching with input biometric information (for example, see Japanese Patent Laid-Open No. 2004-145447) is suggested.
In an example of the known art as described above, a fingerprint matching device selects a registered fingerprint matching an input fingerprint by comparing a selecting parameter value, representing the feature amount of the input fingerprint, with the selecting parameters of a plurality of registered fingerprints, and matches the selected registered fingerprint with the input fingerprint. Then, as the selecting parameters, the rate of a ridge region to the whole fingerprint region, a spacing between a ridge and a valley, the number of weighted feature points, and a value obtained by weighting a ridge direction in a plurality of feature points with the confidence level of the feature points are used.